Stereochemical systematics of metal clusters. Structural

Mar 1, 1978 - Boon-Keng Teo, P. Eisenberger, B. M. Kincaid. J. Am. Chem. Soc. , 1978, 100 (6), pp 1735–1744. DOI: 10.1021/ja00474a015. Publication ...
0 downloads 0 Views 1MB Size
Teo, Eisenberger, Kincaid

/

Structural Characterization of [(q s-C5H5)CoP(C6Hs)2]2+

1735

Stereochemical Systematics of Metal Clusters. Structural Characterization of a Genuine One-Electron Metal-Metal Bonded Cobalt Dimer [(q5-C5H5)CoP(CsH5)2]2+ and Related Compounds by Extended X-Ray Absorption Fine Structure Spectroscopy Boon-Keng Teo,* P. Eisenberger, and B. M. Kincaid Contribution from the Bell Laboratories, Murray Hill, New Jersey 07974. Received June 15, I977

Abstract: Application of extended x-ray absorption fine structure (EXAFS) spectroscopy to the chemically unstable, first "genuine" "one-electron" metal-metal bonded cobalt dimer [CpCoPPh2]2+ provides strong structural evidence that the metal-metal bond weakens significantly upon oxidation from the neutral to the monocationic species. This is consistent only with the theory that the electron being removed comes from an orbital which is highly bonding between the two cobalt atoms. Rather unexpectedly, however, the cobalt-cobalt distance increases by only 0.08 A in going from the neutral compound (2.57 A) to its monocation (2.65 A). It is concluded that the stereochemical influence on metal framework imposed by varying the number of metal cluster electrons decreases drastically in going from antibonding to bonding to (or) nonbonding orbitals (with respect to metal-metal interactions). The metal-ligand and metal-metal (if any) distances of cobaltocene and the series of phosphido-bridged cobalt dimers with the general formula [ C ~ C O P P ~ ~ ] ~ ( O (where H ) , ~m = 0, n = 0, + I ; rn = I,n = + I ; C p = s5-C5H5; and Ph = CsH5) have also been determined by this technique. The weak metal-metal contributions to the EXAFS spectra of the dimeric species necessitate the development of a difference-Fourier technique in the data analysis.

Introduction The stereochemical influence of valence electrons on metal cluster systems has been the subject of many recent studies.'-'O Oxidation(s) of metal-metal nonbonded systems such as cis-[CpFe(CO)PPh2]2" ( n = 0. + 1 , + 2 ) 3 and cis[CpFeYSR]2" (Y = CO, R = Ph, n = 0; Y = CO, R = Me, n = + l ; Y = NCCH3, R = Et, n = +2; Cp = q5-C5H5; Ph = CgH5)4a-Chave been shown to cause a stepwise decrease i n metal-metal distance from a nonbonding (3.498 (4), 3.39 A) in the neutral species to a half-bonding (3.14 (2), 2.925 (4) A) in the monocations to a bonding (2.764 (4), 2.649 (7) A) value in the dications. These tremendous decreases in metal-metal distance only were rationalized by the stepwise removal of two electrons from the highly metal-metal antibonding orbital. In a simplified two-level approximation for these dimers, the number of "metal cluster electrons" (viz., the number of electrons primarily responsible for metal-metal interactions) decreases from four (two in the bonding al, and two in the antibonding b2* orbital, aI2b2**,under CzCsymmetry) to three (al2b2*') to two (a12b2*O) for n = 0, +1, +2 in these iron dimers.2 Reductions of metal-metal systems such as M2(C0)6(PMe2)2" ( M = Fe, Ru; n = 0, -1, -2; Me = CH3) and M~(C0)8(PMe2)2"( M = Cr, Mo, W; n = 0 , -1, -2) have also been r e p ~ r t e d Though .~ no structural information is available, extensive spectroscopic data suggest that the metal-metal distance increases upon stepwise reduction. This is again in accord with the two-level model in that the antibonding metal-metal orbital is successively populated with one and two electrons upon reduction, resulting i n the electronic configurations ofa12b2*0, al2b2*I, and aI2b2**for n = 0, -1, -2 for these latter systems. In contrast, the stereochemical consequences of the oxidation of metal-metal bonded systems are less obvious. The tetramercapto-bridged dimer Cp*Mo2(SMe)4 which has a single Mo-Mo bond has been oxidized by one electron to yield the monocation. Structural characterizations revealed only a small increase in the Mo-Mo distance, going from 2.603 (2) 8, in the neutral to 2.617 (4) A in the monocationic species6 Both of these values are, however, within the bonding range. In view of this uncertainty as well a s the recent controversy 0002-7863/78/ 1.500-1735$01 .OO/O

over the relationship, if any, between metal-metal distance and the degree of metal-metal bonding,' we decided to perform a structural determination on the "genuine" "one-electron'' metal-metal bonded (ai Ib2*O) system [CpCoPPh2]2+ * by extended x-ray absorption fine structure (EXAFS) spectroscopy."-25 This paper reports the analysis of the EXAFS spectra of cobaltocene and a series of phosphido-bridged cobalt dimers with the general formula [CpCoPPh2]2(0H)," (where m = O,n=O,+l;m= 1,n=+1;Cp=q5-C5H~;andPh=CgH5). The neutral compound has a bonding Co-Co distance of 2.56 ( I ) A.9 Attempts to determine the Co-Co distance in the monocation by single-crystal x-ray diffraction have led to the diamagnetic [CpCoPPh2]20Hf cation (owing to its sensitivity toward moisture in the crystal growing process) which has a Co. -Co distance of 2.90 (1) A.8 To our surprise, the EXAFS spectroscopy (described herein) indicates a significant weakening (as determined by a nearly twofold increase in DebyeWaller factor) in the cobalt-cobalt bond upon oxidation but only a relatively small increase in the cobalt-cobalt distance from 2.56 8, in the neutral to 2.65 8, in the monocation.

EXAFS Spectroscopy Two major developments have recently been advanced which make EXAFS a highly attractive structural tool. The first is the formulation and the understanding of the physics involved.' [ - I 6 The oscillatory part of the x-ray absorption rate ( F ) of a particular element beyond its absorption edge (e.g., K edge) is given by

+

sin (2kr; 4 i ( k ) ) (1) r; where FO is the atomic absorption cross section, F ; ( k ) is the backscattering amplitude from each of the Ni neighboring atoms of the ith kind with a Debye-Waller factor u; (which includes both thermal vibration and static disorder) at a distance r, away, and 4 i ( k ) is the phase shift experienced by the X

0 1978 American Chemical Society

1736

Journal of the American Chemical Society

/

100:6

/ March 15, 1978

outgoing and the backscattered photoelectron due to the absorber and the ith scatterer, respectively. The variable k is the photoelectron wavevector which is related to the photon energy E by

where Eo is the energy threshold for an electron ejected from the absorbing atom. It is apparent from eq 1 and 2 that if we know the amplitude Fi(k) and the phase 4 i ( k ) functions we can determine not only the interatomic distances ri but also the Debye-Waller factors ui and, under favorable conditions, the number of atoms Ni. 16-25It should be mentioned that the m 0 Debye- Waller factor as determined by EXAFS spectroscopy 7800 7700 7800 7900 8000 8100 8200 8300 8400 8500 is different from that implied by the conventional crystallogE raphy in that it refers to the root mean square relative vibrab tional amplitude between two atoms along the bond direction and not the absolute root mean square displacement of individual atoms. The second major development is the availability of synchrotron radiation a t Stanford Synchrotron Radiation Laboratory (SSRL) which greatly improves the signal to noise ratio by -1 O5 over the conventional x-ray s o u ~ c e s .The ~ ~ ,real ~~ attractiveness of this technique, however, lies in its structural application to complex biological or chemical systems where single crystals are not available (viz., polymeric or amorphous solid, liquid, solution, gas, etc.). The accuracy of EXAFS structural determination depends on two factors. First, in order to measure interatomic distances and Debye-Waller factors accurately, the phase shift and the amplitude functions (respectively) must be to a large extent insensitive to chemical environment and thus can be transE ferred from one chemical system to another. Indeed this has Figure 1. X-ray absorption spectra (Co K edge) of cobaltocene (a) and been demonstrated to be the case.17c,19,20,25 Secondly, the phase [CpCoPPhzlz (b). 4 ( k ) and the amplitude F ( k ) information must be either predetermined empirically from model compounds17~’9~20-22 pectedly weak cobalt-cobalt bond in the paramagnetic [CpCoPor calculated theoretically from first principles.12-15While Phz] z+PF6-. both approaches have met with significant success, we note that All spectra were run as a compressed neat solid pellet (x X 2 X 40 the recent theory of Lee and Beni14 on EXAFS amplitude and mm3 where the thickness x is typically 1.5 mm) loaded in a drybox phase functions has been shown to provide interatomic disand sealed with thin Kapton tape. The cells were kept under an inert tances accurate to f O . O 1 A and Debye-Waller factors accuatmosphere until just prior to measurements. rate to *IO% (in single distance systems). Subsequent parX-Ray Absorption Measurements. All x-ray absorption measureameterization of these theoretical functions with simple anaments described herein were performed at Stanford Synchrotron Radiation Laboratory (SSRL) using the synchrotron radiation from lytical forms provide a set of working parameters which can Stanford Positron Electron Accelerating Ring (SPEAR).z6 The be used for curve fitting of EXAFS ~ p e c t r a . ~ ~ ~ , ~ ID

Experimental Section Materials. All reactions were carried out under an argon atmosphere. Dichloromethane and toluene were carefully dried and freshly distilled from PzOs before use. The products were stored and handled in a drybox. Cobaltocene (1) was purchased from Strem Chemicals and used without further purification. Following Hayter’s method,1° [CpCoPPhzlz (2) was prepared from the reaction of CpCo(CO)2 with PzPh4 in toluene and purified via recrystallization from toluene. The purity of the product was checked by infrared spectroscopy and elemental analysis. The monocation [CpCoPPh2]2+PF6- (3) was prepared from the quantitative reaction of [CpCoPPhzIz with AgPF6 in freshly prepared, dried, and deoxygenated dichloromethane according to the known method.* The monocation was confirmed by elemental analysis and infrared and electron paramagnetic spectroscopies. The latter exhibits at room temperature the highly characteristic well-resolved 3 I -hyperfine-line EPR spectrums with a g value of 2.035 and an intensity pattern of 1:2:3. . . 1 6 . . . 3:2:1. The monocation [CpCoPPh2]2+PFg- in CHzCIz was then allowed to react with a small amount of water for 20 min. The reaction mixture was subsequently rotavaced to give an unknown mixture of [CpCoPPh2]2+PF6- and [CpCoPPh2]20H+PF6- (4). The known structure* of the diamagnetic [CpCoPPhz]20H+PF6- serves as an “internal” standard for the determination of the distance of the ex-

J

spectrometer used was EXAFS I originally constructed by Kincaid, Eisenberger, and Sayers which has been described el~ewhere.~’ The x-ray beam (slit = I X 20 mm2) from SPEAR, after being monochromated by a channel-cut [220] silicon crystal, passes through first one ionization chamber which measures the incident beam intensity lo, then the sample, and finally another ionization chamber which measures the transmitted intensity 1. Nitrogen was the detecting gas used in both ionization chambers. The EXAFS spectra were typically recorded with an integration time of 2 s/point with 400 steps covering about 800 eV above the edge.

Data Analysis Data Reduction. Figures l a and 1 b show typical plots of px = In ( l o / [ ) as a function of x-ray photon energy E at and above the K edge of cobalt for CpzCo (1) and [CpCoPPhzlz (2), respectively. For the EXAFS analysis, it is necessary to plot the normalized fine structure x ( k ) = ( p po)/po as a function of the photoelectron wavevector

-

k=

d$(E- Eo)

Since we will be using theoretical amplitude and phase functions in the following analysis which necessitate the leastsquares refinement of the energy threshold Eo (viz., an arbitrarily chosen experimental Eo may not be consistent with theoretical phase shift^),'^,*^ we choose a reasonable Eo of

Teo, Eisenberger, Kincaid

/ Structural Characterization of I ( ~ ~ ~ - C S H ~ ) C O P ( C ~ H ~ ) ~ ] ~ +

1737

b

a

4

8

6

IO

12

(Ae’)

k d

I

I

I

I

I

4

6

8

10

k

I 12

1

I

I

I

I

I

4

6

0

10

12

(i-’)

k,

J

(i-l)

Figure 2. Unfiltered (solid curves) and Fourier filtered (dashed curves) , y ( k ) k 3vs. k EXAFS data: (a) Cp2Co; (b) LCpCoPPhzJz; (c) [CpCoPPhz]z+; The ordinate scales are (a) -2.86 to 2.79; (b) -6.40 to 5.88; (c) -9.06 to 10.71; and (d) -7.73 to 6.64 (d) [CpCoPPhz]2+/[CpCoPPh2]20H+. A-3.

7750 eV for all compounds studied. After conversion to k space in A-‘, the data were multiplied by k 3 and the background-a large part of which is due to the ligand absorption as well as the effect that the ionization chamber efficiency changes with photon energy-was removed by using a cubic spline technique (three sections with Ak = 4 A-l each).28aThe x ( k ) k 3 vs. k data were then corrected for the 1.10 dropoff via Victoreen’s true p o / p = CX3 - DX4 equation with C = 141, D = 33.2 for cobalt.29 The resulting curves (solid) are shown in Figures 2(a)-(d). For the purpose of curve fitting, the high frequency noise and the residual background in each spectrum were further removed by a Fourier filtering technique.28bIt involves Fourier transforming the x ( k ) k 3data into R space, selecting the distance ( R ) range to be kept, and back transforming to k space. The Fourier transforms of the data in Figures 2(a)-(d) are depicted in Figures 3(a)-(d), respectively. Smooth filtering windowsof0.8-3.0,0.8-3.2,0.8-3.4,and0.8-3.4A wereused for 1, 2, 3, and 4, respectively. The resulting filtered data, subsequently truncated a t -4 and 12.6 A-l, are compared (dashed vs. solid curves) with the unfiltered data in Figures 2(a)-(d). These filtered data were used for the following curve fitting. Least-Squares Refinements. We use the least-squares minimization technique to fit the filtered spectra with a mul-

tiple-term semiempirical expression of the EXAFS model. A nonlinear least-squares program, which utilizes Marquardt’s scheme for iterative estimation of nonlinear least-squares parameters via a compromise combination of gradient (when far from minimum) and Taylor series (when close to a minimum) method, was used.30 For [CpCoPPh2]2 which contains three different types of nearest neighbors (viz., five cyclopentadienyl carbon, two phosphorus, and one cobalt atoms), a three-term fit of the following expression must be used.

I

x ( k ) k 3= N NCFC(kC)e-20c2kczkC2 X

+

sin ( 2 k m + @ c ( k c ) ) rC2 NpFp(kp)e-2uP2kp2kp2

sin (2kprp + @ P ( ~ P ) ) rP2 NcoFco(kCo)e-2uco2kco2kCo2 X

+

X

sin t2kc0rc0 + @co(kco)) (3) rco2

j

The terms Fi(ki), &(ki), Ni, u,,ri, and ki denote the amplitude, the phase, the number of bonds, the Debye-Waller fac-

Journal of the American Chemical Society

1738

/

100.6

1 March

15, 1978

b

I

C

I

lil d

iA1

Figure 3. Fourier transforms of the unfiltered EXAFS data shown in Figure 2. The ordinate scales are (a) 0- 124; (b) 0-237: (c) 0-290: and (d) 0-309 A-4.

tor, the bond distance, and the photoelectron wavevector, respectively, of the ith type neighboring atoms where i = C , P, Co. For such a complex system, it has been found impractical to vary the overall scale factor N , the amplitude functions, the number of bonds, the Debye-Waller factors, and the distances simultaneously even with a detailed knowledge of phase shifts. The reason is that significant correlations can occur within and between the two sets of functions ( 4 ( k ) ,EO,r] and ( F ( k ) ,u, A'}of each term as well as between different scattering terms. It is obvious that strong correlations can render the individual parameters nonunique and/or result in false minima in curve fitting. Empirically, one can obtain (1) the amplitude function by curve fitting the EXAFS spectrum of a model compound (preferably a single-term system) with a known structure by fixing the coordination number and the Debye-Waller factor (either known or reasonably assumed; in the latter case, all the Debye-Waller factors determined subsequently for the un-

known compounds will be relative to that assumed for the model compound) and (2) the phase shifts by holding the known distance fixed. These functions can then be transferred to unknown systems analyzed in a similar fashion (such as background removal, cutoffs of data, Fourier filtering, etc., so as to minimize systematic errors) in order to determine the number of bonds and/or Debye-Waller factors as well as the interatomic distances. This procedure works reasonably well for most systems. However, for multiatom, multidistance systems with ( 1 ) more than two or three contributions from scatterers of similar (or identical) atomic numbers and similar distances; (2) one term dominating the EXAFS spectrum; (3) unknown bond type and/or bond numbers and unknown Debye-Waller factors, care must be taken to ensure the validity or significance of the resulting parameters. In the first case, trading of the amplitudes between different terms and/or between the amplitudes and the distances can occur. One in-

Teo, Eisenberger, Kincaid

/ Structural Characterization of [AS-C ~ H ~ ) C O P ( C ~ H S ) ~ ] ~ ’ a

Table I. Theoretical Amplitude Parameters A (A), B (A), and C (A-l) and Phase Parameters PO,P I , P2, and P3 for Atom Pairs X-Y = Co-C, Co-P, and Co-Co Where X Is the Absorber and Y Is the Backscatterer

A

B C

PO PI p2 p3

CO-C

co-P

co-co

2.122 0.632 0.88 2.977 -1.415 0.0362 27.0

0.747 0.240 3.26 0.205 -1.345 0.0320 27.0

0.640 0.191 6.66 2.455 - 1.200 0.0253 -59.5

teresting case is the “covering up” of the difference in interatomic distances in a two-term fit by an “apparent larger” Debye-Waller factor in a one-term fit for a one-scatterer two-distance system. This severely limits the resolution of the distance determination to no better than 0.05 A for a data length of ca. Ak = 10 An excellent example is our recent EXAFS analysis of rubredoxin and its model compounds.23 In the second case, it is advantageous to use the difference EXAFS technique in conjunction with Fourier filtering as will be described in this paper (or other variation of it) by first best fitting the dominant term(s) and subtracting it from the observed spectrum followed by (either with or without further Fourier filtering) fitting the difference spectrum with the minor term(s). On the other hand, if the difference i n EXAFS amplitude is not sufficiently significant, it is more accurate to refine each term or terms in stages (while holding others fixed). For (3), it is possible to either (a) assume some reasonable Debye-Waller factors (from vibrational data of analogous compounds) in order to determine the coordination numbers or (b) assume certain bond ratios and search for the x 2 minimum (assuming that the resulting Debye-Waller factors are reasonable). The latter has been successfully applied22 to a number of polymer-bound rhodium catalysts which allows differentiation of various plausible structures based upon the determination of bond ratios which correspond to the minimum X2.

In this paper we employ the recently reported theoretical amplitude and phase functions in the data analysis. These parameterized functions, as well as the number of bonds N , , are held fixed in all our fitting procedures. The parameterized theoretical amplitude and phase functions are

and

where the parameters { A ,B, C) and {PO, P I ,P2, P3}were taken from the l i t e r a t ~ r eand ~ ~listed ~ , ~ in Table 1. Since the phase functions are unique only when a particular energy threshold Eo is specified, our choice of Eo = 7750 eV may not be consistent with the theoretical Eo’s for each of the different types of bonds for which the theoretical phase shift & ( k i ) is defined. We therefore allow a different Eo value for each type of bond via least-squares refining the difference AEoi(eV) = EoexP in where k is the experimental wavevector with EoexP (= 7750 eV) and ki is the theoretical wavevector with Eoith.Thus, for [CpCoPPh2]2, ten parameters are varied: the overall scale factor N , three Debye-Waller factors UC, up, and ucorthree distances rc, rp, and reo, and three threshold energy differ-

1739

-

1

1

X ro 1

I

4

I

I

I

6

8

10

I

12

h (A-’ 1

Figure 4. (a) Theoretical fit (dotted curve) of the filtered EXAFS spectrum (solid curve) for [CpCoPPhzlz; (b) the three backscattering components in the filtered EXAFS spectrum (-) of [CpCoPPh2]2 as resolved by The ordinate scales are -5.93 theory: SC (- - -), 2P (- - -), 1Co (. to 5.33 A-3 for both (a) and (b).

-

e).

ences AEoc, AEop, and AEoc0. The best fit for [CpCoPPhzlz is shown in Figure 4(a). For CpzCo, there is only one kind of neighboring atoms. The model reduces to one term (first term of eq 3) with four parameters i n the least-squares refinement: the overall scale factor N , the Debye-Waller factor UC, the Co-C distances rc, and the threshold energy difference UOC. For the two cationic species 3 and 4, attempts to fit the EXAFS spectra with a three-term model (eq 3) failed to give reasonable oca,reo, and AE0co owing to the relatively small magnitude of the single cobalt contribution (as a scatterer) i n comparison to that of five cyclopentadienyl carbons and two phosphorus atoms. Even in the neutral parent compound 2, the contributions from the carbon and the phosphorus atoms outweigh that of the cobalt atom, as exemplified in Figure 4(b). Upon oxidation, it is expected that the cobalt-cobalt bond weakens significantly. The resulting increase in Debye-Waller factor uc0 and the increase in Co-Co distance, if any, will further diminish the cobalt contribution to the EXAFS. This effect is evident from the Fourier transform of 2,3, and 4 shown in Fi ure 3. For 2, there are two peaks: a stronger one a t r = 1.60 (uncorrected for phase shifts) due to both carbon and phosphorus atoms and a weaker one at r = 2.1 1 8,(uncorrected) due to the cobalt atom. In 3 and 4, the second peak diminishes to about one-half the

‘f

Journal of the American Chemical Society

1740

/

100.6

/ March 15, 1978

a

I

4

I

I

I

I

6

8

10

12

)

lid-? b

4

6

8

12

10

k G-1) c

0

1

4

6

8 k

10

2

3

.

12

6-7

Figure 5. Difference EXAFS spectra (dotted curves) as obtained by subtracting the best fits (dashed curves) with a two-term (5C 2P) model from the filtered data (solid curves): (a) [CpCoPPhzlz; (b) [CpCoPPh2]2+; (c) [CpCoPPhzl2+/LCpCoPPhzJzOH+. The ordinate scales are (a) -5.93 to 5.33; (b) -5.82 to 5.92; and (c) -6.46 to 6.35 A-’.

+

height in 2, indicating a significant weakening of the Co-Co bond.3 To assess the relative strength and the distance of the Co-Co bond in 3 and 4, we resort to the difference Fourier transform technique.32 It involves first fitting the filtered data with a

Figure 6. Fourier transforms of the difference EXAFS spectra shown in Figure 5. The ordinate scales are (a) 0-58; (b) 0-27; and (c) 0-33 A-”

Teo, Eisenberger, Kincaid

/ Structural Characterization of [/q5-C5H5)CoP/C6H5)2]2+

1741 a

two-term model which contains the EXAFS contributions from five carbon atoms and two phosphorus atoms (first two terms of eq 3 ) as shown in Figures S(a)-(c), for 2,3, and 4 (dashed curves), respectively. The model (5C 2P) was then subtracted from the data, resulting in a difference EXAFS spectrum which presumably contains the cobalt contribution and some residual background not removed by Spline technique or filtered by Fourier filtering. These difference spectra are shown also in Figures 5(a)-(c) as dotted curves. The Fourier transforms (Figures 6(a)-(c)) of these difference maps reveal clearly the cobalt contribution(s) as the only discernible feature. Fourier filtering was then performed on these difference spectra with a window of 1.4-3.4 A to remove the residual background and/or residual contributions, if any, from Co-C and Co-P bonds. The resulting filtered difference spectra were then fitted with a one-term (cobalt) model (except 4, which requires two different cobalt distances owing to the ca. 2:l mixture of 3 and 4) as shown in Figures 7(a)-(c). The resulting least-squares refined interatomic distances and the Debye-Waller factors (with estimated errors)33 are tabulated in Table 11.

+

Results It is evident from Table I1 that EXAFS spectroscopy can provide accurate structural parameters for noncrystalline solid. For [CpCoPPhz] 2 and [CpCoPPh2]20H+, which have been structurally characterized by the single-crystal x-ray diffraction method, the agreement is exceedingly good. The Co-C distance of 2.094 (2) A in Cp2Co determined by EXAFS is significantly greater than the Fe-C distance of 2.064 (6) in Cp2Fe34abut substantially smaller than the Ni-C distance of 2.196 (8) A in Cp2Ni,34bin accord with the notion that Cp2Co has one, and Cp2Ni has two, extra electrons over Cp2Fe occupying a metal-carbon antibonding orbital. The two sets of somewhat different bond distances determined from the EXAFS spectrum of [CpCoPPh2]2 agree reasonably well with the reported single-crystal x-ray diffraction values. In particular, the Co-C, Co-P, and Co-Co distances of 2.034 (4), 2.169 (19), and 2.572 (9) A, respectively, obtained by directly fitting the filtered data (cf. Figure 4a), agree very well with the corres onding reported values of 2.046 (20), 2.16 ( l ) , and 2.56 ( I ) .9 On the other hand, the difference-Fourier technique gave Co-C, Co-P (fitting the filtered data with Co-C and Co-P contributions only), and Co-Co (fitting the filtered difference map with Co-Co contribution only) distances of 2.054 (43), 2.1 19 (30), and 2.601 ( I ) A, in somewhat worse agreement with the literature results. This phenomenon is readily understandable in that the Co-Co backscattering amplitude in the neutral compound amounts to about 25% (cf. Figure S(a)) of the overall amplitude and its sinusoidal function (sin (2krc0 4co(k))is not completely orthogonal to those of the carbon and the phosphorus atoms35 such that ignoring it will lead to errors in Co-C and Co-P distances in the initial fitting which subsequently lead to an error in the Co-Co distance in the difference map. For the charged species, the EXAFS contribution from the cobalt scattering is substantially less owing to the significant increase in Debye-Waller factor as a result of the weakening in metal-metal bond (ca 15%; cf. Figures 5(b) and (c)). We expect a better fit with just Co-C and Co-P contributions and therefore more accurate bond distances for these systems via the difference-Fourier technique. For the monocationic species [CpCoPPh2]2+, we determined the Co-C, Co-P, and Co-Co distances to be 2.071 (6), 2.223 (9), and 2.649 (4) A, respectively. For the mixture of [CpCoPPh2]2+ and [CpCoPPh2]zOH+ (ca. 2:l), we found Co-C and Co-P at 2.051 (8) and 2.226 (6) A, respectively, and two Co-Co distances of 2.644 (1) (due to [CpCoPPh2]2+) and

g:

+

4

6

8

10

12

k

Figure 7. Theoretical fits (dotted curves) of the Fourier filtered (solid curves) difference EXAFS spectra from Figures 5 and 6 with cobalt contributions only: (a) [CpCoPPh2]2,single distance; (b) [CpCoPPh2]2+, single distance; (c) [CpCoPPh2]2+/LCpCoPPhz12OH+, two distances. The ordinate scales are (a) -1.29 to 1.28; (b) -0.61 to 0.61; and (c) -0.77 to 0.75 A-3.

2.884 (1) A (due to [CpCoPPh2]20H+). The Co-C, Co-P, and Co-Co distances in the structurally characterized [CpCoPPhz]zOH+ are 2.078,2.213 (6), and 2.901 (5) A, respectively.* It should be mentioned that in our EXAFS anal-

Journal of the American Chemical Society

1142

/

100:6

/ March

15, I978

Table 11. The Least-Squares Refined Interatomic Distances (A) and Debye-Waller Factors (A) with Fitting Errors (Excluding Systematic Errors Such as Background Removal, Fourier Filtering, and Difference-Fourier Analysis] Which Amount to ca. O S % in r and 10% in u) co-c co-P co-co r Q r U r U

2.094 (2) 2.034 (4) 2.054 (43)c 2.071 (6)e 2.051 (8)g

cp2coa

[CpCoPPh2]2' [CpCoPPhzlz [CpCoPPh2]2+ [CpCoPPh2]2OH+

0.0.57 (6)" 0.000 (1 7) 0.004 (53)'~'

0.028 (17)e8' 0.000 (45)g.i

2.169 (19) 2.119 (30)c 2.223 (9)'

0.030 ( 5 5 ) c 0.001 (26)e

2.572 (9) 2.601 ( l ) d 2.649 (4)f

2.226 ( 6 ) g

0.013 (2S)g

2.884

0.072 (15)

2.644 ( l ) h

0.058 (4)

0.066 ( l ) d 0.100 (2)f 0.082 0.095

~~

Fit with data truncated at 5 and 12.6 A-' (cf. Figure 2a). b Fit with a three-term (5C + 2P + 1Co) model (cf. Figure 4a). Fit with a two-term (5C 2P) model (cf. Figure 5a). d Fitting of the difference EXAFS spectrum (cf. Figure 7a) with cobalt contribution only. e Fit with a two-term (5C 2P) model (cf. Figure Sb). f Fitting of the difference EXAFS spectrum (cf. Figure 7b) with cobalt contribution only. g Fit with a two-term (5C + 2P) model (cf. Figure Sc). Fitting of the difference EXAFS spectrum (cf. Figure 7c) with two Co-Co distances. The near-zero values for the Debye-Waller factors of the Co-C bonds is an artifact of the theoretical backscattering amplitude for light scatterers with atomic number